Akash Garg^*

Journal on Artificial Intelligence, Vol.8, pp. 273-298, 2026, DOI:10.32604/jai.2026.079401 - 12 May 2026

Abstract As cyber threats continue to evolve in scale and sophistication, the need for intelligent and adaptive security mechanisms has become increasingly urgent. Intrusion Detection Systems (IDS) are critical components in safeguarding computer networks from malicious activities. This review paper presents a comprehensive analysis of recent advancements in deep learning-based IDS, examining various architectures such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), autoencoders, and generative adversarial networks (GANs). The study compares traditional intrusion detection techniques with modern deep learning approaches, highlighting their strengths, limitations, and suitability for real-world deployment. Special attention is given to… More >

Jianping Wu¹, Nian Li^2,*, Libin Dong³, Ping Zhang⁴

Journal on Artificial Intelligence, Vol.8, pp. 259-271, 2026, DOI:10.32604/jai.2026.075753 - 12 May 2026

Abstract This paper presents a static fusion framework that enhances object detection by integrating camera and LiDAR-based detection results. The proposed method focuses on associating 2D candidate bounding boxes from a camera detector with 3D candidate boxes from a LiDAR detector using an Intersection over Union (IoU)-based matching approach. To enhance the quality of 2D detection, we refine the baseline Cascade R-CNN detector by incorporating a dual self-attention mechanism into both the backbone and the region proposal network (RPN), resulting in the DA-Cascade R-CNN. This enhancement strengthens the network’s ability to detect small or distant objects More >

Harri Renney^1,*, Maxim Nethercott¹, Nathan Renney², Peter Hayes¹

Journal on Artificial Intelligence, Vol.8, pp. 231-257, 2026, DOI:10.32604/jai.2026.078487 - 17 April 2026

Abstract This paper provides systemisation on the emerging design patterns and paradigms for Large Language Model (LLM)-enabled multi-agent systems (MAS), evaluating their practical utility across various domains, bridging academic research and industry practice. We define key architectural components, including agent orchestration, communication mechanisms, and control-flow strategies, and demonstrate how these enable rapid development of modular, domain-adaptive solutions. Three real-world case studies are tested in controlled, containerised pilots in telecommunications security, national heritage asset management, and utilities customer service automation. Initial empirical results show that, for these case studies, prototypes were delivered within two weeks and pilot-ready More >

Kadriye Simsek Alan^*, Busra Senel Kahyaoglu

Journal on Artificial Intelligence, Vol.8, pp. 203-230, 2026, DOI:10.32604/jai.2026.078359 - 15 April 2026

Abstract Cardiovascular diseases remain one of the leading causes of mortality worldwide, making early and reliable diagnosis a critical challenge for modern healthcare systems. In this study, a systematic comparative performance analysis of widely used machine learning algorithms is conducted for the early detection of heart disease using tabular clinical data. Rather than proposing a novel model architecture, the primary objective is to provide a fair, reproducible, and clinically meaningful evaluation of commonly adopted classifiers under consistent experimental conditions. The Kaggle Heart Failure dataset is employed, and multiple machine learning models—including tuned Random Forest, tuned XGBoost,… More >

Daniil Kamakaev, Khaled Mahbub^*

Journal on Artificial Intelligence, Vol.8, pp. 183-202, 2026, DOI:10.32604/jai.2026.077823 - 15 April 2026

Abstract This paper presents an innovative system designed to automate the analysis of candidate interviews by integrating multiple analytical techniques into a single multimodal framework. This system combines text sentiment analysis, audio sentiment analysis, keyword extraction, and Mel-Frequency Cepstral Coefficients (MFCC) feature extraction to evaluate candidate performance holistically. This system employs text sentiment analysis using VADER and transformer-based sentiment features (probability-based outputs), audio sentiment analysis with an SVM model trained on both IEMOCAP and MELD datasets, keyword extraction via KeyBERT, and audio feature extraction including MFCCs, delta MFCCs, pitch, and energy to evaluate candidate performance holistically. More >

Long Li, Hengyang Wu^*, Na Wang

Journal on Artificial Intelligence, Vol.8, pp. 169-182, 2026, DOI:10.32604/jai.2026.077462 - 23 March 2026

Abstract Zero-shot learning as an emerging approach in image captioning techniques, has garnered significant attention from researchers in recent years due to its ability to accomplish tasks without requiring specific category training data. Existing zero-shot image captioning schemes largely rely on traditional language models, which exhibit low efficiency and suboptimal generation quality. To address this issue, this study proposes Hamiltonian Monte Carlo for Image Captioning (HMCIC). This method first models the image captioning task as a probabilistic sampling problem in parameter space, integrating semantic matching and syntactic coherence into an energy function to guide the generation… More >

Haoyang Xu^*

Journal on Artificial Intelligence, Vol.8, pp. 153-167, 2026, DOI:10.32604/jai.2026.078014 - 11 March 2026

Abstract This paper conducted a comprehensive study on the robustness of three widely used DFD deep learning models—namely, ResNet50, FreqNet, and Xception v1—to controlled perturbation attacks and frequency masking across a range of 12 different distortions. The study was performed on 254,166 ForenSynth test images, characterizing the distribution of FSI-drop values derived from over 3.05 million paired predictions. The distribution of FSI-drop values is sharply peaked around zero: 99.7% of the samples exhibit |Δ| < 0.1, and the maximum |Δ| ≈ 1.5 × 10⁻³, indicating high baseline stability. In terms of perturbation-wise comparison, Gaussian blur dominates, yielding… More >

Syed Nisar Hussain Bukhari^*, Sehar Altaf

Journal on Artificial Intelligence, Vol.8, pp. 137-152, 2026, DOI:10.32604/jai.2026.077943 - 11 March 2026

Abstract Knee osteoarthritis is a progressive degenerative joint disorder that leads to pain, stiffness, and reduced mobility, significantly affecting quality of life. Early and reliable diagnosis is essential for effective disease management, yet conventional radiographic assessment remains time-consuming and subject to inter-observer variability. This study presents a comparative deep learning (DL) based approach for automated severity classification of knee osteoarthritis using plain radiographic images. Multiple pretrained convolutional neural network architectures, including EfficientNetB3, InceptionNet, VGG19, ResNet, and EfficientNetV2S, were evaluated within a transfer learning paradigm. All models were trained and assessed on a publicly available dataset to More >

A. S. M. Masudur Rahman¹, Muhammad Zunair Zamir¹, Syed Sajid Ullah^2,*, Salman Khan², Maria Saman¹, Naqash Bahadar¹

Journal on Artificial Intelligence, Vol.8, pp. 119-136, 2026, DOI:10.32604/jai.2026.076274 - 24 February 2026

Abstract Vehicle detection plays a pivotal role in autonomous driving, traffic monitoring, and intelligent surveillance systems. While YOLOv8 offers strong real-time performance, its detection accuracy is often limited by insufficient feature stability and suboptimal multi-scale feature fusion in complex scenes. To address these issues, we propose an enhanced YOLOv8 framework that retains the original backbone and detection head for efficiency while introducing targeted improvements to the neck architecture. Specifically, the model incorporates an Exponential Moving Average (EMA) feature layer to stabilize learning through temporally smoothed feature representations, which reduces noise and enhances generalization, and integrates GhostConv… More >

Jeevan Pralhad Tonde^*, Satish Sankaye

Journal on Artificial Intelligence, Vol.8, pp. 107-118, 2026, DOI:10.32604/jai.2026.075857 - 13 February 2026

Abstract As education continues to evolve alongside artificial intelligence, there is growing interest in how large language models (LLMs) can support more personalized and intelligent learning experiences. This study focuses on building a domain-specific question answering (QA) system tailored to computer science education, with a particular emphasis on Java programming. While transformer-based models such as BERT, RoBERTa, and DistilBERT have demonstrated strong performance on general-purpose datasets like SQuAD, they often struggle with technical educational content where annotated data is scarce. To address this challenge, we developed a custom dataset, JavaFactoidQA, consisting of 1000 fact-based question–answer pairs… More >